[mirror_edk2.git] / MdePkg / Library / BaseLib / Ipf / setjmp.s

/// @file\r
///  Contains an implementation of setjmp and longjmp for the\r
///  Itanium-based architecture.\r
///\r
/// Copyright (c) 2006, Intel Corporation                                                         \r
/// All rights reserved. This program and the accompanying materials                          \r
/// are licensed and made available under the terms and conditions of the BSD License         \r
/// which accompanies this distribution.  The full text of the license may be found at        \r
/// http://opensource.org/licenses/bsd-license.php                                            \r
///                                                                                           \r
/// THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,                     \r
/// WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.             \r
/// \r
/// Module Name: setjmp.s\r
///\r
///\r
\r
  .file  "setjmp.s"\r
\r
#include  "asm.h"\r
#include  "ia_64gen.h"\r
\r
/// int SetJump(struct jmp_buffer save)\r
///\r
///  Setup a non-local goto.\r
///\r
/// Description:\r
///\r
///  SetJump stores the current register set in the area pointed to\r
///  by "save".  It returns zero.  Subsequent calls to "LongJump" will\r
///  restore the registers and return non-zero to the same location.\r
///\r
/// On entry, r32 contains the pointer to the jmp_buffer\r
///\r
\r
PROCEDURE_ENTRY(SetJump)\r
  //\r
  //  Make sure buffer is aligned at 16byte boundary\r
  //\r
  mov    r32 = r33        \r
\r
    add     r10 = -0x10,r0  ;;  // mask the lower 4 bits\r
    and     r32 = r32, r10;; \r
    add     r32 = 0x10, r32;;   // move to next 16 byte boundary\r
\r
    add    r10 = J_PREDS, r32  // skip Unats & pfs save area\r
  add    r11 = J_BSP, r32\r
  //\r
  //  save immediate context\r
  //\r
  mov    r2 = ar.bsp     // save backing store pointer\r
  mov    r3 = pr         // save predicates\r
  ;;\r
  //\r
  // save user Unat register\r
  //\r
  mov    r16 = ar.lc        // save loop count register\r
  mov    r14 = ar.unat     // save user Unat register\r
\r
  st8    [r10] = r3, J_LC-J_PREDS\r
  st8    [r11] = r2, J_R4-J_BSP\r
  ;;\r
  st8    [r10] = r16, J_R5-J_LC\r
  st8    [r32] = r14, J_NATS  // Note: Unat at the \r
                    // beginning of the save area\r
  mov    r15 = ar.pfs\r
  ;;\r
  //\r
  //  save preserved general registers & NaT's\r
  //\r
  st8.spill  [r11] = r4, J_R6-J_R4\r
  ;;\r
  st8.spill  [r10] = r5, J_R7-J_R5 \r
    ;;\r
  st8.spill  [r11] = r6, J_SP-J_R6\r
  ;;\r
  st8.spill  [r10] = r7, J_F3-J_R7 \r
    ;;\r
  st8.spill  [r11] = sp, J_F2-J_SP\r
  ;;\r
  //\r
  // save spilled Unat and pfs registers\r
  //\r
  mov    r2 = ar.unat       // save Unat register after spill\r
  ;;\r
  st8    [r32] = r2, J_PFS-J_NATS  // save unat for spilled regs\r
  ;;\r
  st8    [r32] = r15          // save pfs\r
  //\r
  //  save floating registers \r
  //\r
  stf.spill  [r11] = f2, J_F4-J_F2\r
  stf.spill  [r10] = f3, J_F5-J_F3 \r
  ;;\r
  stf.spill  [r11] = f4, J_F16-J_F4\r
  stf.spill  [r10] = f5, J_F17-J_F5 \r
  ;;\r
  stf.spill  [r11] = f16, J_F18-J_F16\r
  stf.spill  [r10] = f17, J_F19-J_F17 \r
  ;;\r
  stf.spill  [r11] = f18, J_F20-J_F18\r
  stf.spill  [r10] = f19, J_F21-J_F19 \r
  ;;\r
  stf.spill  [r11] = f20, J_F22-J_F20\r
  stf.spill  [r10] = f21, J_F23-J_F21 \r
  ;;\r
  stf.spill  [r11] = f22, J_F24-J_F22\r
  stf.spill  [r10] = f23, J_F25-J_F23 \r
  ;;\r
  stf.spill  [r11] = f24, J_F26-J_F24\r
  stf.spill  [r10] = f25, J_F27-J_F25 \r
  ;;\r
  stf.spill  [r11] = f26, J_F28-J_F26\r
  stf.spill  [r10] = f27, J_F29-J_F27 \r
  ;;\r
  stf.spill  [r11] = f28, J_F30-J_F28\r
  stf.spill  [r10] = f29, J_F31-J_F29 \r
  ;;\r
  stf.spill  [r11] = f30, J_FPSR-J_F30\r
  stf.spill  [r10] = f31, J_B0-J_F31    // size of f31 + fpsr\r
  //\r
  // save FPSR register & branch registers\r
  //\r
  mov    r2 = ar.fpsr  // save fpsr register\r
  mov    r3 = b0 \r
  ;;\r
  st8    [r11] = r2, J_B1-J_FPSR\r
  st8    [r10] = r3, J_B2-J_B0\r
  mov    r2 = b1\r
  mov    r3 = b2 \r
  ;;\r
  st8    [r11] = r2, J_B3-J_B1\r
  st8    [r10] = r3, J_B4-J_B2\r
  mov    r2 = b3\r
  mov    r3 = b4 \r
  ;;\r
  st8    [r11] = r2, J_B5-J_B3\r
  st8    [r10] = r3\r
  mov    r2 = b5 \r
  ;;\r
  st8    [r11] = r2\r
  ;;\r
  //\r
  // return\r
  //\r
  mov    r8 = r0        // return 0 from setjmp\r
  mov    ar.unat = r14  // restore unat\r
  br.ret.sptk  b0\r
\r
PROCEDURE_EXIT(SetJump)\r
\r
\r
//\r
// void LongJump(struct jmp_buffer *)\r
//\r
//  Perform a non-local goto.\r
//\r
// Description:\r
//\r
//  LongJump initializes the register set to the values saved by a\r
//  previous 'SetJump' and jumps to the return location saved by that\r
//  'SetJump'.  This has the effect of unwinding the stack and returning\r
//  for a second time to the 'SetJump'.\r
//\r
\r
PROCEDURE_ENTRY(LongJump)\r
  //\r
  //  Make sure buffer is aligned at 16byte boundary\r
  //\r
  mov    r32 = r33        \r
\r
    add     r10 = -0x10,r0  ;;  // mask the lower 4 bits\r
    and     r32 = r32, r10;; \r
    add     r32 = 0x10, r32;;   // move to next 16 byte boundary\r
\r
  //\r
  // caching the return value as we do invala in the end\r
  //\r
///  mov    r8 = r33          // return value\r
  mov    r8 = 1              // For now return hard coded 1\r
\r
  //\r
  //  get immediate context\r
  //\r
  mov    r14 = ar.rsc    // get user RSC conf \r
  add    r10 = J_PFS, r32  // get address of pfs\r
  add    r11 = J_NATS, r32\r
  ;;\r
  ld8    r15 = [r10], J_BSP-J_PFS  // get pfs\r
  ld8    r2 = [r11], J_LC-J_NATS      // get unat for spilled regs\r
  ;;\r
  mov    ar.unat = r2\r
  ;;\r
  ld8    r16 = [r10], J_PREDS-J_BSP  // get backing store pointer\r
  mov    ar.rsc = r0        // put RSE in enforced lazy \r
  mov    ar.pfs = r15\r
  ;;\r
  \r
  //\r
  // while returning from longjmp the BSPSTORE and BSP needs to be\r
  // same and discard all the registers allocated after we did\r
  // setjmp. Also, we need to generate the RNAT register since we\r
  // did not flushed the RSE on setjmp.\r
  //\r
  mov    r17 = ar.bspstore  // get current BSPSTORE\r
  ;;\r
    cmp.ltu  p6,p7 = r17, r16  // is it less than BSP of \r
(p6)  br.spnt.few  .flush_rse\r
  mov    r19 = ar.rnat    // get current RNAT\r
  ;;\r
  loadrs              // invalidate dirty regs\r
  br.sptk.many  .restore_rnat    // restore RNAT\r
\r
.flush_rse:\r
  flushrs\r
  ;;\r
  mov    r19 = ar.rnat    // get current RNAT\r
  mov    r17 = r16      // current BSPSTORE\r
  ;;\r
.restore_rnat:\r
  //\r
  // check if RNAT is saved between saved BSP and curr BSPSTORE\r
  //\r
  dep    r18 = 1,r16,3,6   // get RNAT address\r
  ;;\r
  cmp.ltu  p8,p9 = r18, r17  // RNAT saved on RSE\r
  ;;\r
(p8)  ld8    r19 = [r18]    // get RNAT from RSE\r
  ;;\r
  mov    ar.bspstore = r16  // set new BSPSTORE  \r
  ;;\r
  mov    ar.rnat = r19    // restore RNAT\r
  mov    ar.rsc = r14    // restore RSC conf\r
\r
\r
  ld8    r3 = [r11], J_R4-J_LC    // get lc register\r
  ld8    r2 = [r10], J_R5-J_PREDS   // get predicates\r
  ;;\r
  mov    pr = r2, -1\r
  mov    ar.lc = r3\r
  //\r
  //  restore preserved general registers & NaT's\r
  //\r
  ld8.fill  r4 = [r11], J_R6-J_R4\r
  ;;\r
  ld8.fill  r5 = [r10], J_R7-J_R5 \r
  ld8.fill  r6 = [r11], J_SP-J_R6\r
  ;;\r
  ld8.fill  r7 = [r10], J_F2-J_R7\r
  ld8.fill  sp = [r11], J_F3-J_SP\r
  ;;\r
  //\r
  //  restore floating registers \r
  //\r
  ldf.fill  f2 = [r10], J_F4-J_F2\r
  ldf.fill  f3 = [r11], J_F5-J_F3 \r
  ;;\r
  ldf.fill  f4 = [r10], J_F16-J_F4\r
  ldf.fill  f5 = [r11], J_F17-J_F5 \r
  ;;\r
  ldf.fill  f16 = [r10], J_F18-J_F16\r
  ldf.fill  f17 = [r11], J_F19-J_F17\r
  ;;\r
  ldf.fill  f18 = [r10], J_F20-J_F18\r
  ldf.fill  f19 = [r11], J_F21-J_F19\r
  ;;\r
  ldf.fill  f20 = [r10], J_F22-J_F20\r
  ldf.fill  f21 = [r11], J_F23-J_F21\r
  ;;\r
  ldf.fill  f22 = [r10], J_F24-J_F22\r
  ldf.fill  f23 = [r11], J_F25-J_F23 \r
  ;;\r
  ldf.fill  f24 = [r10], J_F26-J_F24\r
  ldf.fill  f25 = [r11], J_F27-J_F25\r
  ;;\r
  ldf.fill  f26 = [r10], J_F28-J_F26\r
  ldf.fill  f27 = [r11], J_F29-J_F27\r
  ;;\r
  ldf.fill  f28 = [r10], J_F30-J_F28\r
  ldf.fill  f29 = [r11], J_F31-J_F29 \r
  ;;\r
  ldf.fill  f30 = [r10], J_FPSR-J_F30\r
  ldf.fill  f31 = [r11], J_B0-J_F31 ;;\r
\r
    //\r
  // restore branch registers and fpsr\r
  //\r
  ld8    r16 = [r10], J_B1-J_FPSR  // get fpsr\r
  ld8    r17 = [r11], J_B2-J_B0    // get return pointer\r
  ;;\r
  mov    ar.fpsr = r16\r
  mov    b0 = r17\r
  ld8    r2 = [r10], J_B3-J_B1\r
  ld8    r3 = [r11], J_B4-J_B2\r
  ;;\r
  mov    b1 = r2\r
  mov    b2 = r3\r
  ld8    r2 = [r10], J_B5-J_B3\r
  ld8    r3 = [r11]\r
  ;;\r
  mov    b3 = r2\r
  mov    b4 = r3 \r
  ld8    r2 = [r10]\r
  ld8    r21 = [r32]      // get user unat\r
  ;;\r
  mov    b5 = r2\r
  mov    ar.unat = r21\r
\r
  //\r
  // invalidate ALAT\r
  //\r
  invala ;;\r
\r
  br.ret.sptk  b0\r
PROCEDURE_EXIT(LongJump)\r
\r
\r
Commit	Line	Data
878ddf1f	1	/// @file\r
	2	/// Contains an implementation of setjmp and longjmp for the\r
	3	/// Itanium-based architecture.\r
	4	///\r
	5	/// Copyright (c) 2006, Intel Corporation \r
	6	/// All rights reserved. This program and the accompanying materials \r
	7	/// are licensed and made available under the terms and conditions of the BSD License \r
	8	/// which accompanies this distribution. The full text of the license may be found at \r
	9	/// http://opensource.org/licenses/bsd-license.php \r
	10	/// \r
	11	/// THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS, \r
	12	/// WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED. \r
	13	/// \r
	14	/// Module Name: setjmp.s\r
	15	///\r
	16	///\r
	17	\r
	18	.file "setjmp.s"\r
	19	\r
	20	#include "asm.h"\r
	21	#include "ia_64gen.h"\r
	22	\r
	23	/// int SetJump(struct jmp_buffer save)\r
	24	///\r
	25	/// Setup a non-local goto.\r
	26	///\r
	27	/// Description:\r
	28	///\r
	29	/// SetJump stores the current register set in the area pointed to\r
	30	/// by "save". It returns zero. Subsequent calls to "LongJump" will\r
	31	/// restore the registers and return non-zero to the same location.\r
	32	///\r
	33	/// On entry, r32 contains the pointer to the jmp_buffer\r
	34	///\r
	35	\r
	36	PROCEDURE_ENTRY(SetJump)\r
	37	//\r
	38	// Make sure buffer is aligned at 16byte boundary\r
	39	//\r
	40	mov r32 = r33 \r
	41	\r
	42	add r10 = -0x10,r0 ;; // mask the lower 4 bits\r
	43	and r32 = r32, r10;; \r
	44	add r32 = 0x10, r32;; // move to next 16 byte boundary\r
	45	\r
	46	add r10 = J_PREDS, r32 // skip Unats & pfs save area\r
	47	add r11 = J_BSP, r32\r
	48	//\r
	49	// save immediate context\r
	50	//\r
	51	mov r2 = ar.bsp // save backing store pointer\r
	52	mov r3 = pr // save predicates\r
	53	;;\r
	54	//\r
	55	// save user Unat register\r
	56	//\r
	57	mov r16 = ar.lc // save loop count register\r
	58	mov r14 = ar.unat // save user Unat register\r
	59	\r
	60	st8 [r10] = r3, J_LC-J_PREDS\r
	61	st8 [r11] = r2, J_R4-J_BSP\r
	62	;;\r
	63	st8 [r10] = r16, J_R5-J_LC\r
	64	st8 [r32] = r14, J_NATS // Note: Unat at the \r
65	// beginning of the save area\r
66	mov r15 = ar.pfs\r
67	;;\r
68	//\r
69	// save preserved general registers & NaT's\r
70	//\r
71	st8.spill [r11] = r4, J_R6-J_R4\r
72	;;\r
73	st8.spill [r10] = r5, J_R7-J_R5 \r
74	;;\r
75	st8.spill [r11] = r6, J_SP-J_R6\r
76	;;\r
77	st8.spill [r10] = r7, J_F3-J_R7 \r
78	;;\r
79	st8.spill [r11] = sp, J_F2-J_SP\r
80	;;\r
81	//\r
82	// save spilled Unat and pfs registers\r
83	//\r
84	mov r2 = ar.unat // save Unat register after spill\r
85	;;\r
86	st8 [r32] = r2, J_PFS-J_NATS // save unat for spilled regs\r
87	;;\r
88	st8 [r32] = r15 // save pfs\r
89	//\r
90	// save floating registers \r
91	//\r
92	stf.spill [r11] = f2, J_F4-J_F2\r
93	stf.spill [r10] = f3, J_F5-J_F3 \r
94	;;\r
95	stf.spill [r11] = f4, J_F16-J_F4\r
96	stf.spill [r10] = f5, J_F17-J_F5 \r
97	;;\r
98	stf.spill [r11] = f16, J_F18-J_F16\r
99	stf.spill [r10] = f17, J_F19-J_F17 \r
100	;;\r
101	stf.spill [r11] = f18, J_F20-J_F18\r
102	stf.spill [r10] = f19, J_F21-J_F19 \r
103	;;\r
104	stf.spill [r11] = f20, J_F22-J_F20\r
105	stf.spill [r10] = f21, J_F23-J_F21 \r
106	;;\r
107	stf.spill [r11] = f22, J_F24-J_F22\r
108	stf.spill [r10] = f23, J_F25-J_F23 \r
109	;;\r
110	stf.spill [r11] = f24, J_F26-J_F24\r
111	stf.spill [r10] = f25, J_F27-J_F25 \r
112	;;\r
113	stf.spill [r11] = f26, J_F28-J_F26\r
114	stf.spill [r10] = f27, J_F29-J_F27 \r
115	;;\r
116	stf.spill [r11] = f28, J_F30-J_F28\r
117	stf.spill [r10] = f29, J_F31-J_F29 \r
118	;;\r
119	stf.spill [r11] = f30, J_FPSR-J_F30\r
120	stf.spill [r10] = f31, J_B0-J_F31 // size of f31 + fpsr\r
121	//\r
122	// save FPSR register & branch registers\r
123	//\r
124	mov r2 = ar.fpsr // save fpsr register\r
125	mov r3 = b0 \r
126	;;\r
127	st8 [r11] = r2, J_B1-J_FPSR\r
128	st8 [r10] = r3, J_B2-J_B0\r
129	mov r2 = b1\r
130	mov r3 = b2 \r
131	;;\r
132	st8 [r11] = r2, J_B3-J_B1\r
133	st8 [r10] = r3, J_B4-J_B2\r
134	mov r2 = b3\r
135	mov r3 = b4 \r
136	;;\r
137	st8 [r11] = r2, J_B5-J_B3\r
138	st8 [r10] = r3\r
139	mov r2 = b5 \r
140	;;\r
141	st8 [r11] = r2\r
142	;;\r
143	//\r
144	// return\r
145	//\r
146	mov r8 = r0 // return 0 from setjmp\r
147	mov ar.unat = r14 // restore unat\r
148	br.ret.sptk b0\r
149	\r
150	PROCEDURE_EXIT(SetJump)\r
151	\r
152	\r
153	//\r
154	// void LongJump(struct jmp_buffer *)\r
155	//\r
156	// Perform a non-local goto.\r
157	//\r
158	// Description:\r
159	//\r
160	// LongJump initializes the register set to the values saved by a\r
161	// previous 'SetJump' and jumps to the return location saved by that\r
162	// 'SetJump'. This has the effect of unwinding the stack and returning\r
163	// for a second time to the 'SetJump'.\r
164	//\r
165	\r
166	PROCEDURE_ENTRY(LongJump)\r
167	//\r
168	// Make sure buffer is aligned at 16byte boundary\r
169	//\r
170	mov r32 = r33 \r
171	\r
172	add r10 = -0x10,r0 ;; // mask the lower 4 bits\r
173	and r32 = r32, r10;; \r
174	add r32 = 0x10, r32;; // move to next 16 byte boundary\r
175	\r
176	//\r
177	// caching the return value as we do invala in the end\r
178	//\r
179	/// mov r8 = r33 // return value\r
180	mov r8 = 1 // For now return hard coded 1\r
181	\r
182	//\r
183	// get immediate context\r
184	//\r
185	mov r14 = ar.rsc // get user RSC conf \r
186	add r10 = J_PFS, r32 // get address of pfs\r
187	add r11 = J_NATS, r32\r
188	;;\r
189	ld8 r15 = [r10], J_BSP-J_PFS // get pfs\r
190	ld8 r2 = [r11], J_LC-J_NATS // get unat for spilled regs\r
191	;;\r
192	mov ar.unat = r2\r
193	;;\r
194	ld8 r16 = [r10], J_PREDS-J_BSP // get backing store pointer\r
195	mov ar.rsc = r0 // put RSE in enforced lazy \r
196	mov ar.pfs = r15\r
197	;;\r
198	\r
199	//\r
200	// while returning from longjmp the BSPSTORE and BSP needs to be\r
201	// same and discard all the registers allocated after we did\r
202	// setjmp. Also, we need to generate the RNAT register since we\r
203	// did not flushed the RSE on setjmp.\r
204	//\r
205	mov r17 = ar.bspstore // get current BSPSTORE\r
206	;;\r
207	cmp.ltu p6,p7 = r17, r16 // is it less than BSP of \r
208	(p6) br.spnt.few .flush_rse\r
209	mov r19 = ar.rnat // get current RNAT\r
210	;;\r
211	loadrs // invalidate dirty regs\r
212	br.sptk.many .restore_rnat // restore RNAT\r
213	\r
214	.flush_rse:\r
215	flushrs\r
216	;;\r
217	mov r19 = ar.rnat // get current RNAT\r
218	mov r17 = r16 // current BSPSTORE\r
219	;;\r
220	.restore_rnat:\r
221	//\r
222	// check if RNAT is saved between saved BSP and curr BSPSTORE\r
223	//\r
224	dep r18 = 1,r16,3,6 // get RNAT address\r
225	;;\r
226	cmp.ltu p8,p9 = r18, r17 // RNAT saved on RSE\r
227	;;\r
228	(p8) ld8 r19 = [r18] // get RNAT from RSE\r
229	;;\r
230	mov ar.bspstore = r16 // set new BSPSTORE \r
231	;;\r
232	mov ar.rnat = r19 // restore RNAT\r
233	mov ar.rsc = r14 // restore RSC conf\r
234	\r
235	\r
236	ld8 r3 = [r11], J_R4-J_LC // get lc register\r
237	ld8 r2 = [r10], J_R5-J_PREDS // get predicates\r
238	;;\r
239	mov pr = r2, -1\r
240	mov ar.lc = r3\r
241	//\r
242	// restore preserved general registers & NaT's\r
243	//\r
244	ld8.fill r4 = [r11], J_R6-J_R4\r
245	;;\r
246	ld8.fill r5 = [r10], J_R7-J_R5 \r
247	ld8.fill r6 = [r11], J_SP-J_R6\r
248	;;\r
249	ld8.fill r7 = [r10], J_F2-J_R7\r
250	ld8.fill sp = [r11], J_F3-J_SP\r
251	;;\r
252	//\r
253	// restore floating registers \r
254	//\r
255	ldf.fill f2 = [r10], J_F4-J_F2\r
256	ldf.fill f3 = [r11], J_F5-J_F3 \r
257	;;\r
258	ldf.fill f4 = [r10], J_F16-J_F4\r
259	ldf.fill f5 = [r11], J_F17-J_F5 \r
260	;;\r
261	ldf.fill f16 = [r10], J_F18-J_F16\r
262	ldf.fill f17 = [r11], J_F19-J_F17\r
263	;;\r
264	ldf.fill f18 = [r10], J_F20-J_F18\r
265	ldf.fill f19 = [r11], J_F21-J_F19\r
266	;;\r
267	ldf.fill f20 = [r10], J_F22-J_F20\r
268	ldf.fill f21 = [r11], J_F23-J_F21\r
269	;;\r
270	ldf.fill f22 = [r10], J_F24-J_F22\r
271	ldf.fill f23 = [r11], J_F25-J_F23 \r
272	;;\r
273	ldf.fill f24 = [r10], J_F26-J_F24\r
274	ldf.fill f25 = [r11], J_F27-J_F25\r
275	;;\r
276	ldf.fill f26 = [r10], J_F28-J_F26\r
277	ldf.fill f27 = [r11], J_F29-J_F27\r
278	;;\r
279	ldf.fill f28 = [r10], J_F30-J_F28\r
280	ldf.fill f29 = [r11], J_F31-J_F29 \r
281	;;\r
282	ldf.fill f30 = [r10], J_FPSR-J_F30\r
283	ldf.fill f31 = [r11], J_B0-J_F31 ;;\r
284	\r
285	//\r
286	// restore branch registers and fpsr\r
287	//\r
288	ld8 r16 = [r10], J_B1-J_FPSR // get fpsr\r
289	ld8 r17 = [r11], J_B2-J_B0 // get return pointer\r
290	;;\r
291	mov ar.fpsr = r16\r
292	mov b0 = r17\r
293	ld8 r2 = [r10], J_B3-J_B1\r
294	ld8 r3 = [r11], J_B4-J_B2\r
295	;;\r
296	mov b1 = r2\r
297	mov b2 = r3\r
298	ld8 r2 = [r10], J_B5-J_B3\r
299	ld8 r3 = [r11]\r
300	;;\r
301	mov b3 = r2\r
302	mov b4 = r3 \r
303	ld8 r2 = [r10]\r
304	ld8 r21 = [r32] // get user unat\r
305	;;\r
306	mov b5 = r2\r
307	mov ar.unat = r21\r
308	\r
309	//\r
310	// invalidate ALAT\r
311	//\r
312	invala ;;\r
313	\r
314	br.ret.sptk b0\r
315	PROCEDURE_EXIT(LongJump)\r
316	\r
317	\r